The evolution, propagation, and spread of flash drought in the Central United States during 2012

The United States (US) drought of 2012 was an extreme event of historic proportions. At its peak, it covered most of the contiguous US and represented the fourth largest drought by aerial extent since 1895 (Knapp et al 2015). Within the Great Plains region of the Central US, precipitation deficits during the most critical portion of the growing season exceeded those of the most extreme drought years during the Dust Bowl (Schubert et al 2004), and were the greatest noted in the modern-day historical record dating to 1895 (Hoerling et al 2014). Severe impacts were observed on regional ecosystems (Mallya et al 2013, Knapp et al 2015, Sippel et al 2016, Wolf et al 2016, Zhou et al 2017) and agricultural productivity (Boyer et al 2013); in the heavily irrigated region of central Nebraska, the drought led to increased water usage and subsequent groundwater declines that took years to recover (Young et al 2018). In total, economic losses from the drought were estimated to be in excess of $30 billion across the entire nation (NCEI 2017). Impacts from the 2012 drought ultimately effected energy systems, created water shortages and navigational disruptions, and reduced employment in rural areas (Grigg 2014).

Drought in the Central US has been attributed to numerous local to global factors including surface-atmosphere feedbacks, persistent synoptic patterns, and large-scale teleconnections (Basara et al 2013). As such, the overall evolution of the 2012 drought was inherently complex. It was embedded within an extended period of continuous drought in the US that spanned late 2010 through mid 2015, a period that also included notable extreme drought events in the Southern Great Plains (Hoerling et al 2013, Seager et al 2013) and California (Griffin and Anchukaitis 2014, Seager et al 2015). In addition, La Nina conditions during the winter of 2011–2012, resulted in broad regions of the US experiencing drier-than-normal conditions (Rippey 2015). The resulting implications for the regional atmospheric circulation, which included a displaced stormtrack, limited moisture return, and broad subsidence, led to large negative standardized anomalies in precipitation, runoff, and soil moisture and positive anomalies in temperature (Hoerling et al 2013). Persistent negative soil moisture anomalies were also observed (Otkin et al 2016), which were further strongly coupled to atmospheric processes during this period (Basara and Christian 2018).

The nature of the 2012 Central US drought's evolution is of particular interest, specifically its rapid geographic expansion and marked intensification from non-drought to extreme/exceptional drought conditions during the growing season (AghaKouchak 2014, McEvoy et al 2016, Otkin et al 2016). As noted by Otkin et al (2016) and displayed in figure 1, the vast majority of the Central US domain (the region east of the Rocky Mountains and west of the Great Lakes) was drought-free or abnormally dry, D0, as late as mid-May 2012 as defined by the US Drought Monitor (Svoboda et al 2002). However, drought quickly enveloped the region, with large portions of the Central US deteriorating into D3 (Extreme) or D4 (Exceptional) drought by July. During this period of rapid intensification, drought intensities increased in magnitude by four or even five US Drought Monitor categories in fewer than two months.

Zoom In Zoom Out Reset image size

Such rapid intensification is consistent with the definition of flash drought by Otkin et al (2018). Flash droughts occur during periods when extreme precipitation deficits combined with increased surface temperature, wind speed, downwelling shortwave radiation, and/or vapor pressure deficits persist for several weeks. This leads to the rapid depletion of soil moisture, increased evaporative stress, and the rapid intensification of drought conditions and associated impacts. Christian et al (2019) developed a methodology to identify flash drought using a standardized form of the ratio between evapotranspiration and potential evapotranspiration: the standardized evaporative stress ratio (SESR). This ratio is based on previous satellite-based studies focused on evaporative stress (Anderson et al 2007a, 2007b, Otkin et al 2013) and is inversely proportional to the amount of evaporative stress on the environment. The use of SESR is advantageous for flash drought identification as it incorporates numerous variables (Otkin et al 2014, Ford et al 2015, Hobbins et al 2016, McEvoy et al 2016, Otkin et al 2016, Otkin et al 2019.

Because flash drought is a relatively new area of study and yet has critical impacts, the objectives of this study were to (1) quantify the rapid evolution and expansion of the 2012 historic drought in the Central US, using the SESR, and (2) identify critical feedbacks via land-atmosphere interactions that accentuated the 2012 drought's intensification, perpetuation, and propagation.

The North American Regional Reanalysis (NARR) was used to investigate the evolution of the 2012 flash drought across the Central US, in particular to examine land-atmosphere interactions that develop and propagate of flash drought conditions. The NARR has a spatial resolution of 32 km with 29 vertical levels (Mesinger et al 2006). Computed quantities include the SESR, as well as the convective triggering potential (CTP) and low level humidity index (HI) methodology originally developed by Findell and Eltahir (2003a). The domain for analysis was bounded by a north and south latitude of 49°N and 33.5°N, respectively, and a west and east longitude of −104°W and 87°W, respectively.

Using the flash drought definition of Otkin et al (2018) and the methodology developed by Christian et al (2019), NARR grid points across the Central US in 2012 were analyzed for flash drought occurrence via standardized ESR (SESR) and changes in SESR. ESR is defined as:

$$\begin{eqnarray*}&&{\rm{E}}{\rm{S}}{\rm{R}}=\displaystyle \frac{{\rm{E}}{\rm{T}}}{{\rm{P}}{\rm{E}}{\rm{T}}},\end{eqnarray*}$$

where ET is evapotranspiration and PET potential evapotranspiration. Specifically within the Noah land surface model in the NARR dataset, total surface ET is computed from soil evaporation, transpiration from the vegetation canopy, evaporation of dew/frost or canopy-intercepted precipitation, and snow sublimation (Mesinger et al 2006) while PET is calculated using the modified Penman scheme from Mahrt and Ek (1984). SESR and the change in SESR are defined as:

$$\begin{eqnarray*}&&{\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{ijp}=\displaystyle \frac{{\rm{E}}{\rm{S}}{{\rm{R}}}_{ijp}-\overline{{\rm{E}}{\rm{S}}{{\rm{R}}}_{ijp}}}{{\sigma }_{{\rm{E}}{\rm{S}}{{\rm{R}}}_{ijp}}}\end{eqnarray*}$$

$$\begin{eqnarray*}&&{\left({\rm{\Delta }}{\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{{i}jp}\right)}_{z}=\displaystyle \frac{{\rm{\Delta }}{\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{{i}jp}-\overline{{\rm{\Delta }}{\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{{i}jp}}}{{\sigma }_{{\rm{\Delta }}{\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{{i}jp}}},\end{eqnarray*}$$

where ${\rm{S}}{\rm{E}}{\rm{S}}{{\rm{R}}}_{ijp}$ (referred to as SESR) and ${\left({\rm{\Delta }}SES{R}_{ijp}\right)}_{z}$ (referred to as ΔSESR) are the z-scores of ESR and the change in SESR, respectively, for a specific pentad (p) at a specific grid point (i, j). As described in Christian et al (2019), four restrictive criteria were employed to identify flash drought occurrence at a given location. These include two criteria that emphasized the rapid intensification toward drought and two criteria focused on vegetative impacts, summarized as:

• A negative trend of SESR for a minimum length of five pentads,
• A final SESR value below the 20th percentile,
• Pentad-to-pentad changes in SESR (ΔSESR) below the 40th percentile,
• Change in SESR over the entirety of the flash drought below the 25th percentile.

This methodology applied to the NARR dataset has been shown to compare well with the satellite-based evaporative stress index (ESI), drought depiction from the USDM, and previously identified flash drought cases (Christian et al 2019). For locations identified as having experienced flash drought, the date whereby flash drought began was partitioned by month. In addition, the spatial coverage of flash drought across the region was calculated from the aggregate of grid points that experienced flash drought.

Convection is a critical contributor to total precipitation during the growing season across the Central US (Haberlie and Ashley 2019). As such, processes that suppress convective initiation and propagation are critical to drought and flash drought development. Further, the Central US is a region known for strong land-atmosphere interactions (Koster et al 2004, Guo and Dirmeyer 2013, Koster et al 2016) which can yield feedback processes that enhance drought development (Hoerling et al 2013, Roundy et al 2013, Basara and Christian 2018, Wakefield et al 2019). Gerken et al (2018) utilized the CTP and HI framework in the analysis of 2017 flash drought conditions in the Northern Plains of the US and noted that incorporating land-atmosphere interactions has broad utility to understanding flash drought development. As such, to examine feedback processes and convective suppression related to flash drought development and expansion in 2012, the CTP and HI framework was utilized; the framework is described in Findell and Eltahir (2003a, 2003b) whereby CTP is determined by:

• (1)  
  Locating the moist adiabat which intersects the temperature profile 100 hPa above ground level (AGL).
• (2)  
  Integrating the area between this moist adiabat and the temperature profile from 100 hPa AGL to 300 hPa AGL.

Further, HI is found by summing the dewpoint depressions at 50 and 150 hPa AGL as follows:

$$\begin{eqnarray*}&&{\rm{HI}}={\left({T}-{\rm{Td}}\right)}_{50{\rm{hPa}}{\rm{AGL}}}+{\left({T}-{\rm{Td}}\right)}_{150{\rm{hPa}}{\rm{AGL}}}.\end{eqnarray*}$$

The framework quantifies pre-existing instability within the lower portions of the atmosphere, pre-existing moisture in the atmosphere before development of the convective boundary layer, and indirectly the overall atmospheric moisture content in the lower troposphere. Following the methods of Wakefield et al (2019), daily standardized anomalies of CTP and HI were computed for the NARR dataset (1979 through 2017). These standardized anomalies (hereafter, CTPz and HIz) allow for comparison of CTP and HI in both time and space where significant variations in mean CTP and HI values occur. When averaged over longer time periods, these values yield persistence of moisture within the boundary layer as well as covariability with variables that impact evaporative stress (Wakefield et al 2019).

3.1. Flash drought expansion

The flash drought in 2012 began during April, with an epicenter located over north-central Kansas (figure 2) immediately adjacent to an area of long-term drought (which lingered from 2011) across southwest Kansas (figure 1). During May 2012, rapid drought development expanded north into Nebraska, South Dakota, and North Dakota, eastward across Missouri and Illinois, and southeastward into Oklahoma and Arkansas. Conversely, localized above-normal precipitation (positive values of the one-month standardized precipitation index; SPI; McKee et al 1993, 1995) for the month of May across the northern Great Plains and Midwest acted to inhibit flash drought development in southeast Nebraska and Iowa (figure 3). Overall, the areal coverage identified to have experienced flash drought across the central United States in May 2012 spanned approximately 299 000 km (figure 2).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

During June 2012, flash drought expanded in multiple areas across the central Great Plains and Midwest. The first of these areas included southeast Kansas and southwest Oklahoma, located along the periphery of a region that had already experienced flash drought during May. In addition, a broad region extending across southeast Nebraska, Iowa, Illinois, and the southern portion of Wisconsin experienced flash drought during the month of June. In total, these locations contributed approximately 361 000 km² of new land area that experienced rapid onset of drought, the highest among all months of the 2012 growing season.

The remainder of the 2012 growing season saw modest additional flash drought expansion in the Central US. In July 2012, the primary regions exhibiting new flash drought included northern Iowa, southern Minnesota, western Wisconsin, central Missouri, and southern Arkansas, while during August 2012, portions of Oklahoma, Iowa, Minnesota, and North Dakota also experienced flash drought. Flash drought development during September was isolated to the Midwest across northwestern Illinois and several locations in Wisconsin, as isolated areas or 'gaps' within the spatial coverage of flash drought development were finally filled.

The total land area across the Central US study domain that experienced flash drought at any point during the 2012 growing season was approximately 1008 640 km², constituting over 40% of the total land area in the domain. From a historical perspective, based on the Christian et al (2019) methodology applied to the entire 1979–2016 NARR period, the 2012 flash drought ranks as the 2nd highest in total spatial coverage within the study domain during the growing season. While some locations in the analysis and domain did not reach the flash drought criteria as defined by Christian et al (2019), these locations did experience drought onset.

Representative locations within the domain were chosen to illustrate the temporal evolution of flash drought using SESR during different expansion periods within the 2012 growing season (figure 4; locations shown in figure 5). Near the epicenter of the 2012 flash drought event, the overall change in SESR was exceptionally dramatic with values declining 3.5 standard deviations in only 30 d. Specifically, SESR decreased by 2.4 standard deviations during the last 15 d of the rapid drought intensification, which was the largest negative change in SESR for that grid point at that given time of year in the climatological record of NARR (figure 4(a)). For regions to the southeast (north-central Oklahoma) and east (north-central Missouri) of the flash drought epicenter in north-central Kansas, rapid drought development began approximately one week later (figures 4(b) and (c)). Conversely, evaporative stress increased rapidly in central Nebraska in late May 2012, approximately four weeks after rapid onset drought began in Oklahoma and Kansas (figure 4(d)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In west-central Iowa, SESR began to decline in early May 2012, but was moderated by precipitation in May and June (figures 3 and 4(e)). In fact, June 2012 precipitation in the region was at or slightly above normal (figure 3). However, this precipitation was not sufficient to prevent rapid intensification toward drought due to exceptionally hot and cloud-free conditions. As such, SESR sharply declined from late June to mid-July 2012 during a subsequent period of negative precipitation anomalies.

As noted in figure 2, an extensive region of additional flash drought development occurred in late July 2012 for southeastern Minnesota and west-central Wisconsin. This can be seen in figure 4(f), where rapid drought intensification lasted for approximately 40 days and ended in late August 2012. Examination of the SPI reveals that the region received above-normal totals through the early portion of the season which initially staved off drought development (figure 3). However, with the onset of negative precipitation anomalies, flash drought rapidly developed soon thereafter.

3.2. Surface-atmosphere interactions and feedbacks

Not all drought is the same, as fundamentally, various physical processes can inhibit precipitation thus leading to drought at multiple spatial and temporal scales. Overall, the 2012 Central US drought can be considered an historic and extreme event. The rapid intensification and expansion of drought in the Central US occurred at shorter timescales than other locations impacted by the same broad, long-term drought (e.g. Southwest US). This rapid development was consistent with the characteristics of a flash drought, one that began with an epicenter in the central Great Plains followed by quasi-radial expansion throughout the larger Central US domain. While the 2012 drought as a whole has been well established as an historic event, the flash drought component observed in the Central US was also historic, with over 1250 000 km² meeting the criteria for flash drought. This would rank near the top of total flash drought coverage in a year across the Central US domain during the satellite record.

Two main results were revealed from the analysis of flash drought development and extent. First, the Central US domain had near-normal or above-normal SESR values just before the rapid development of drought. Thus, antecedent surface conditions did not serve to inhibit the development of flash drought. Second, the time series and spatial analysis of SESR demonstrated that flash drought did not develop uniformly across the region. Rather, the majority of flash drought in the Central US occurred along areas adjacent to prior flash drought development earlier in the study period. Therefore, the Central US flash drought should not be considered a single, linear event, but rather as an evolving phenomenon that propagated and spread through space and time.

Additionally, the timing of flash drought onset was closely tied to the advection of a hostile air mass from regions already impacted by drought. CTPz and HIz provide information concerning the atmospheric component of land-atmosphere feedbacks and, in the case of the 2012 drought, they demonstrated anomalies of moisture in the lower atmosphere along with the subsequent persistence and movement of the dry air mass. Thus, land-atmosphere feedbacks through the boundary-layer contributed to anomalously dry air (elevated HIz) over drought impacted areas in the Central US. As such, while large-scale dynamics yielded overall environmental conditions conducive for drought development, the downstream advection of hostile air from adjacent, drought impacted areas (1) increased the atmospheric demand, (2) elevated the evaporative stress, all while (3) further suppressing deep convection critical to overall precipitation and water availability of the region. In the end, the timing of increased evaporative stress and positive CTPz and HIz values not only drove flash drought development on a local scale, but accentuated the propagation of flash drought to adjacent areas, eventually spreading across the majority of the Central US.

The 2012 drought in the Central US was a function of many critical land-atmosphere components. While this study focused on rapid drought intensification through time and space, along with the impacts of local and non-local feedbacks associated with land-atmosphere interactions, it is important to note that large-scale drivers of drought were also extremely important (Hoerling et al 2013, Rippey 2015, Otkin et al 2016). In fact, it was the combination and interaction of various physical processes at multiple timescales that led to the cascading evolution of flash drought, sustained drought, and drought impacts during the 2012 growing season.

It is also critical to note that the methodology developed by Christian et al (2019) provides an innovative framework for examining flash drought given its incorporation of multiple environmental conditions within a standardized approach. As such, it should be utilized in drought monitoring as well as future studies that examine additional past drought events in the Central US (e.g. 1988, 2006, etc) along with historical drought development across the globe to better quantify the role and mechanisms of flash drought and land-atmosphere feedbacks in cascading events.

This work was supported, in part, by the NOAA Climate Program Office's Sectoral Applications Research Program (SARP) grant NA130AR4310122, the Agriculture and Food Research Initiative Competitive Grant No. 2013-69002 from the USDA National Institute of Food and Agriculture, the USDA National Institute of Food and Agricultural (NIFA) Grant No. 2016-68002-24967, the USDA-ARS Southern Great Plains Climate Hub Cooperative Agreement 58-3070-8-012, and the NASA SMD Earth Science Division through the Earth Science Application: Water Resources Program. The data from the NARR used in this study is available at https://esrl.noaa.gov/psd/data/gridded/data.narr.html. The authors would also like to acknowledge the anonymous reviewers who provided critical feedback that improved this manuscript.